Unlock real-time access for AI agents with SQL++ and the Model Context Protocol (MCP)
As autonomous AI agents become more powerful and pervasive, developers need ways to securely and reliably connect these agents to operational and analytical data. Whether you’re building customer service agents, intelligent dashboards, or agentic workflows, data access is the backbone of intelligence.
That’s why we’re excited to announce that Couchbase is now officially supported in the Google MCP Toolbox for Databases.
This integration empowers developers to accelerate the development of agentic AI applications by bridging the gap between Couchbase’s high-performance, flexible NoSQL database and the MCP standard for tool-based agent orchestration. By leveraging Couchbase within the MCP Toolbox, teams can unlock intelligent data-driven workflows that are secure, scalable, and context-aware. Developers no longer need to spend time building custom connectors or managing complex access logic. Now they can focus on building smarter agents that query and act on operational data with minimal friction. This integration brings the performance, flexibility, and real-time capabilities of Couchbase to the emerging ecosystem of Model Context Protocol (MCP)-compatible AI tools.
Why Couchbase + MCP Toolbox matters for developers
The Google MCP Toolbox for Databases acts as an MCP server, allowing AI agents to interact with structured data through declarative tool definitions, without the need to expose databases directly or write custom integration code for each use case.
By adding Couchbase support, developers gain:
-
- Rich document modeling with sub-document access and flexible schemas.
- Built-in scalability and sync for edge, mobile, and distributed applications.
- MCP compatibility, enabling AI agents to reason over and query Couchbase data via secure, standardized tool interfaces.
Now, agents can issue natural-language queries that get converted into SQL++ tool calls behind the scenes, without hardcoding logic, creating custom APIs, or exposing database credentials.
How it works
The Google MCP Toolbox for Databases is an open-source MCP server that connects AI orchestration layers (like LangChain, LangGraph, or Anthropic’s Claude) to database-backed tools.
To configure Couchbase access, you define your data sources and tools in a YAML config file:
Sample tools.yaml
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 |
sources: cb-orders: kind: couchbase connectionString: couchbase://localhost:8091 bucket: orders_bucket scope: orders username: toolbox_user password: $CB_PASSWORD tools: get-customer-orders: description: Retrieve recent orders for a specific customer by email. source: cb-orders query: | SELECT order_id, total, status, order_date FROM orders_bucket WHERE customer.email = $email ORDER BY order_date DESC LIMIT 10; parameters: - name: email type: string description: customer email |
Once deployed, your AI agents can invoke get-customer-orders by passing the email parameter, and the MCP Toolbox will translate the request into a secure SQL++ query against Couchbase.
Agentic use cases for Couchbase + MCP
Conversational BI dashboards
Build agents that generate dynamic dashboards and metrics by querying Couchbase for business insights. Couchbase’s SQL++ support allows complex aggregations and JOINs, even across flexible schemas.
-
-
- Example: “What were the top 10 product categories last quarter by revenue?”
-
E-Commerce Agent Assistants
Serve product recommendations, inventory checks, and order histories from JSON documents in Couchbase. Support fast lookups and full-text search to guide users in real-time.
-
-
- Example: “Show me available trail running shoes under $120 in size 11.”
-
IoT and Edge Analytics
Couchbase Mobile and Sync Gateway bring real-time data from edge devices into Couchbase. Agents can analyze telemetry, spot anomalies, or initiate remediation workflows using SQL++ queries.
-
-
- Example: “List sensors with temperature fluctuations over 10 degrees in the past hour.”
-
Secure Customer Copilots
Build personalized agents that help users understand their own data—finances, activity, healthcare, etc.—by securely scoping SQL++ queries to their identity context.
-
-
- Example: “Summarize my recent transactions and categorize them.”
-
LangChain + Couchbase via MCP: integration example
Here’s how you can wire Couchbase tools into a LangChain agent via MCP Toolbox:
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 |
from langgraph.prebuilt import create_react_agent from langchain_openai import ChatOpenAI MCP_ENDPOINT = "http://localhost:8080" tool_box = ToolboxClient(MCP_ENDPOINT) toolbox_tools = toolbox_client.load_toolset("toolset_name") llm = ChatOpenAI( model="gpt-4o", temperature=0.1, api_key=settings.openai_api_key ) agent = create_react_agent( tools=toolbox_tools, model=llm, debug=True ) agent.invoke( {"messages": [{"role": "user", "content": "Get the last few orders for alice@example.com."}]} ) |
Built for production
The MCP Toolbox for Databases offers production-grade features out of the box:
-
- Config-driven deployment with zero downtime
- Connection pooling and retry logic
- OAuth2 / OIDC auth for secure access
- OpenTelemetry support for metrics and traces
- Support for multiple databases from a single config
You can deploy the toolbox as a microservice alongside your orchestration layer (e.g. in GKE, Cloud Run, or Kubernetes Anywhere).
Get started
-
- Google MCP Toolbox for Databases (GitHub)
- Couchbase Developer Portal
- Deploy locally or in the cloud with Docker, Cloud Run, or Kubernetes
In summary
With this new integration, Couchbase joins the growing list of production-ready backends in the Google MCP ecosystem. Developers now have a fast, scalable, and secure way to enable AI agents to access NoSQL document data using standardized protocols.
This lowers the barrier to building intelligent applications that react in real-time to user inputs, contextual data, and live system states, all powered by Couchbase and governed by the open MCP standard.